Skip to main content

Installing dstack

dstack is an open-source orchestrator for running AI workloads across clouds and on-prem infrastructure.

Install and run dstack​

Install and run dstack in your environment (local machine, VM, or cluster for the control plane). For full instructions, see the dstack documentation.

Make sure the dstack API is reachable from the machine running the Transformer Lab API. For example, running dstack with --host 0.0.0.0 and exposing the default port (3000) allows access from other machines.

Gather connection details​

You will need these values when adding dstack as a compute provider in Transformer Lab:

  • server_url — e.g. http://<host>:3000
  • api_token — Bearer token from your dstack server
  • dstack_project — typically main

Once dstack is running, continue to Add dstack as a Compute Provider →